Powering Ethernet, Part 2: Optimizing power consumption during standby operation
Editor’s note: In Part 1 we looked at designing for low power consumption in operation. Now let’s look at standby operation.
As devices become faster and denser, power consumption continues to rise. Studies have shown that standby power can account for up to 25 percent of the electricity consumed in homes. Ethernet is a dominating technology in today’s smart digital home, but the IEEE never considered power consumption as a significant factor in the original specifications. However, as we will show here, with proper analysis and configuration, Energy Efficient Ethernet is achievable.
Ethernet Power Consumption (Standby)
Power consumption during the standby operation is also of concern, especially when designing to specifications such as the ‘One Watt Initiative.’ This is an energy-saving initiative by the International Energy Agency to reduce standby power-use by any appliance to not more than one watt in 2010, and 0.5 watts in 2013, which has given rise to regulations in many countries and regions and impacts device design. Some considerations follow to optimize power savings for Ethernet devices.
Power Down Mode
Typically, the Ethernet device power-down state is controlled via an internal register bit, which, when enabled, completely powers down the device except for the management interface. One example of recent innovation in this space is Micrel’s KSZ8081/91 Ethernet PHY family, which reduces the current consumption further with the ability to slow down the internal clock oscillator during power-down mode. Current consumption can be reduced from around 2.6mA to 1.3mA under these conditions.
Power Saving Mode – Energy Detect
Power saving mode is used to reduce the PHY power consumption automatically when the cable is unplugged or link partner disabled. The receive circuit detects the presence or absence of a signal, commonly known as ‘energy detect’ to enter or exit power saving mode. For example, the KSZ8863/73 switch family supports four modes of operation configured by the Power Management Event Control Register (PMECR):
- Normal Operation Mode is the default mode where all internal blocks are powered and fully functional.
- Power Saving Mode is when the cable is unplugged; the receiver block is powered down, leaving only the energy detect circuitry alive. The transmitter block, all PLL clocks and MAC are enabled. Internal register values will not change, and host interface is ready for CPU access. Once activity is present, the link is established and normal operation resumes.
- Energy Detect Mode is an enhanced version of the power saving mode described above, wherein the device will save more than 50 percent power, relative to normal operation. Here, the, ‘no energy’ detection period prior to entering the low power mode is also configurable. During energy detect mode all internal blocks are disabled, except for the energy detect circuitry. The PHY will also transmit continuous 120ns width pulses at a rate of one pulse per second, rather than the typical (energy consuming) idle pattern.
- Soft Power Down Mode is a mode in which all functional blocks are disabled except for the host interface. Internal register values will not change during this mode. Any host access will wake-up the device from soft power down mode to normal operation mode. A hardware power down pin is also provided to reduce power consumption further, by disabling the host interface.
Table 1. Status of Internal Functional Blocks During Power Management Modes
What about Wake-on-LAN?
Wake-on-LAN (WoL) seems to offer a solution to wake up the system during a low power idle state. However, the reality is that such this feature rarely achieves its goal. In WoL, a special wake up sequence is sent to the Ethernet device, which it detects and asserts an interrupt signal used to notify the host to power up the rest of the system. WoL has arguably never become popular because it is not standards-driven, nor does it have a single common defined wake up sequence, which hinders interoperability across vendors.
A second limitation when minimizing standby power is found in the implementation of WoL functionality in the MAC, not PHY layer. The PHY device is purely a transceiver and is transparent to the received packet contents. It is often not appreciated that both the MAC and PHY must operate in normal and not a low power mode prior to system wake up. As we have shown above, even with no traffic, the idle state of a 100Base-TX PHY consumes full power. Furthermore, most embedded applications, such as IP STB and VoIP phones, already embed the Ethernet MAC inside the core processor, neglecting WoL support. Without hardware support, now the processor too would have to be powered to detect WoL packets. To overcome such limitations of the implementation of WoL Micrel offers such functionality in their latest PHY transceivers, KSZ8091 (10/100BASE-TX) and KSZ9031 (10/100/1000BASE-T).
Achieving Energy Efficient Ethernet—IEEE 802.3az
The IEEE has clearly recognized power consumption inefficiency within Ethernet circuits. Its 802.3az task force, also known as Energy Efficient Ethernet, was targeted to reduce power consumption during periods of low link utilization (idle time). As we have already seen, typical Ethernet traffic utilization of a link is extremely low, estimated to be sub-3 percent for 10/100Base-TX links and even lower for GigE. To achieve this goal, changes are needed to the hardware; however, they must be fully backwards compatible.
Reducing power consumption during low link utilization periods allows for drastic improvement in power efficiency. Known as Low Power Idle (LPI), this technique will disable parts of the PHY transceiver that are not necessary, whilst still maintaining the link integrity. When new frames arrive, the PHY is awoken and the returns to the normal active state. During LPI, a periodic refresh symbol is sent out by the PHY to ensure that the receiver is synchronized. An example of IEEE 802.3az Energy Efficient Ethernet operation is show in Figure 1.
Figure 1: Energy Efficient Ethernet Example (IEEE 802.3az)
As we see here, when a frame arrives for transmission and the link is in the low power mode, it has to wait until the link is awoken before transmission can begin. This does introduce an additional latency Tw_PHY in the data path. Proposed wake up times for 100Base-TX and 1000-BaseT are 30µs and 16.5µs, respectively. Several 100Base-TX (full duplex), 1000Base-T and 10G PHYs are available on the market supporting the IEEE 802.3az standard including KSZ8091 (10/100BASE-TX) and KSZ9031 (10/100/1000BASE-T) PHY transceivers.
Conclusion
IEEE802.3az Energy Efficient Ethernet will prove to be a significant aid in reducing idle period power. Complimenting Energy Efficient Ethernet, additional power savings can also be made both during normal traffic and link down. Although calculating the power consumption of an Ethernet circuit is not straightforward, optimizing power savings for Ethernet devices is achievable.
About the author
Mike Jones has over 25 years of experience in high-tech design in the semiconductor industry. Jones is currently based in Newbury, U.K., where he is Product Marketing Director, responsible for Micrel’s Automotive and Industrial LAN Solutions. Prior to Micrel, Jones worked for several high tech companies, in various engineering roles. including as principal engineer at BT and Fijitsu Telecommunications where he gained more than a decade in design experience in SONET, SDH and PDH systems. Jones also held position of senior FAE and Product Marketing Manager with Micrel for fourteen years prior to my current position.
Jones graduated in 1990 with a 1st in class honors degree in Electric Systems Engineering at Aston University in Birmingham, UK.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
